Hillnote

Cloud-based productivity tools hold your notes hostage in proprietary databases — readable only through their interface, gone the moment you cancel. Hillnote takes the opposite position: every document, database, kanban board, canvas, and slide is stored as plain markdown on your local disk, readable by any text editor, and never dependent on a remote service to open. AI assistance runs on-device through Apple Intelligence or Ollama for inline edits and rephrasing, with auto-mode escalating to a frontier model when the task warrants it — all configurable, all without forcing a cloud round-trip for small tasks.

The differentiating feature is agent readiness baked into the file format itself. Because the workspace is plain markdown on disk, external agents — the vendor specifically names Claude Code, Cowork, Codex, and any MCP client — can read and edit your files directly without exports or proprietary APIs. A built-in STDIO MCP server exposes the workspace to those clients, and a CLI brings the same access inside Hillnote. This is not a bolt-on integration; it is a consequence of the file architecture.

Hillnote fits writers, developers, and privacy-focused individuals who want AI assistance without surrendering file ownership, and who already use or plan to use external agents on their workspace. It breaks as a team tool: the vendor describes no concurrent editing, no shared permissions model, and no collaborative workspace features. A solo practitioner building a personal knowledge base alongside agentic tools is the clear target; a product team looking to replace a shared wiki is not.

Sync across Mac, Windows, iOS, and Android is available and described as opt-in — files stay local unless you turn it on. The workspace can also be extended with HTML tools and npm packages, and published as a doc, blog, or site with a custom domain. The API is not exposed for external programmatic access outside the MCP pathway.